Inference
EN
InternScience

Agents-A1-4B

Agents-A1-4B is a reasoning model from InternScience in the Agents-A1 family. 8 benchmarks count toward its index, in 4 categories.

  • Open weights
  • Reasoning
  • Input: text
  • Released 2026-07
RankedEnough results for an index.
IndexAll categories, out of 100.46.8±9.7Median 51.0
CoverageIndex weight that has results.60%
SpeedOutput tokens per second.—Median 67/s
First tokenSeconds to the first output token.—Median 1.3 s
Input / 1MUS dollars per 1M input tokens.—Median $0.60
Output / 1MUS dollars per 1M output tokens.—Median $2.55
ContextMaximum tokens in one request.262KMedian 262K
EloLMArena rating and rank.—Median 1422

Run locally

4.21B parameters
16-bit
9.8 GB
8-bit
5.7 GB
4-bit
3.7 GB
Weights
Hugging Face (opens in a new tab) ↗

Memory need with a short context. The 4-bit copy fits in 8 GB. Compare local models

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.3 counted54.1107 / 218
CodingCode writing and repair.2 counted33.9194 / 209
ReasoningLogic problems and puzzles.1 counted42.1—
MultimodalTasks with images and text.None counted——
KnowledgeFacts and expert knowledge.None counted——
MultilingualTasks in many languages.None counted——
InstructionTasks with strict prompt rules.2 counted53.828 / 63
MathMath problems.None counted——

Results

8 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
Instruction-Following EvalInstruction94.8%5 / 70
VITA-BenchAgentic40.3%6 / 13
τ²-Bench Tool-Agent-User EvaluationAgentic78.2%72 / 149
Instruction Following BenchmarkInstruction69.1%41 / 57
BrowseCompAgentic66.8%46 / 63
LongBench v2Reasoning52.1%20 / 25
LiveCodeBench v6Coding59.6%48 / 59
Scientific Code BenchmarkCoding29.6%31 / 35
General AI AssistantsAgentic95.1%—
FrontierScience ResearchKnowledge33.3%—
MLE-Bench LiteAgentic22.7%—

Sources

BenchLMCC BY-NC 4.0

History

5 releases
Index47.2 → 46.8
29 Oct 202612 Nov 2026

Questions

How good is Agents-A1-4B?
Its index is 46.8 out of 100. That is place 171 of 283 ranked models.
What is Agents-A1-4B best at?
Instruction: 53.8 out of 100, place 28 of 63.
What is the context window of Agents-A1-4B?
262K tokens.
Is Agents-A1-4B open source?
Yes. Its weights are published.
When was Agents-A1-4B released?
July 2026.

Badge

Agents-A1-4B on Inference360: place 171 of 283, index 46.8
HTML
<a href="https://inference360.ai/models/agents-a1-4b"><img src="https://inference360.ai/badges/agents-a1-4b.svg" alt="Agents-A1-4B on Inference360: place 171 of 283, index 46.8" height="20"></a>
Markdown
[![Agents-A1-4B on Inference360: place 171 of 283, index 46.8](https://inference360.ai/badges/agents-a1-4b.svg)](https://inference360.ai/models/agents-a1-4b)

Compare with

Sage-A1.560.5

More from InternScience

All models
Agents-A1-F16-GGUFUnrankedAgents-A1-FP8UnrankedAgents-A1-Q4_K_M-GGUFUnrankedAgents-A1-Q8_0-GGUFUnrankedAgents-A1Unranked